Internet-Based Electric Meter with Theft Detection, Theft Notification and Consumption Monitoring for Residential Power Lines Using Wireless Network Technology

International Journal of Electrical and Electronic Engineering & Telecommunications Vol. 8, No. 5, September 2019

Maria Criselda B. Loyola, Jhamel B. Bueno, and Roshmeir D. De Leon Electrical Engineering Department, Malayan Colleges Laguna, Cabuyao City, Philippines Email: [email protected]; {jhamelbueno1996; rxmeir.deleon}@gmail.com

Abstract—In support of a proposed bill in the Philippines, electrocution. These are punishable by law under the the Recoverable System Loss Act, which aims to reduce the Republic Act 7832, the “Anti-Electricity, Electric Power system losses cap thus making electricity cheaper, the Lines, Equipment, and Materials Pilferage Act.” objective of this project is to help the distribution utility and Electric pilferage is accounted for under non- the consumer in decreasing the incidences of power theft. technical system loss. Distribution utilities charge each An electric meter which features theft detection and notification, and an internet-based energy consumption consumer a certain percentage for system loss. monitoring system were designed to alleviate the problem. Therefore, even though a consumer does not steal Theft detection is achieved through the utilization of electricity, they are charged higher than what they microcontrollers and current sensors, while theft actually consumed. The distribution utility does not have notification is achieved through the utilization of LoRaWAN full responsibility of the incurred costs due to technology. The theft detection module can identify illegal electricity pilferage, which makes them unmotivated tapping and meter bypassing. A star network of electric perhaps to prioritize addressing the issue. The meters was implemented to test the LoRaWAN in an energy Recoverable System Loss Act proposed by the chairman metering application. The theft detection and notification of the Senate Committee on Energy aims to reduce modules were 100% successful in identifying and transmitting the correct type of theft. The watt-hour the cost of electricity in the Philippines by further measurement of the electric meter exhibited 99.9% accuracy reducing the system losses cap from 8.5% to 5% for in reference to a commissioned digital electric meter. In private distribution utilities and 13% to 10% for coordination with the electric meter, theft notifications and electric cooperatives, and the exemption from value- meter data were accessible to both the consumer and added tax of the system loss charge. distribution utility through an online energy monitoring Given the current situation of the Philippines’ system: enermon.tech. The LoRaWAN gateway was able to electrical distribution sector, the objective of this receive theft notifications from an electric meter node as far project was to design an electric power meter and an as 601 m line of sight with an average round trip time of internet-based energy consumption monitoring system. 33.09 milliseconds in the implemented area. This system provides immediate notification of the presence of pilferage The electric power meter features theft detection for for the distribution utility and consumer to address.  residential power lines, and a theft notification system for both the consumer and distribution utility. This Index Terms—electricity pilferage, illegal tapping, electric power meter may help in improving the way the LoRaWAN, meter bypassing, non-technical system loss, government and distribution utilities address electricity wireless network pilferage in the Philippines. If adapted nationwide, incidences of power theft due to ‘jumpers’ could be I. INTRODUCTION reduced. The electric power meter cannot be tampered, One of the major concerns of Meralco, the country’s which eliminates the occurrences of theft due to meter largest private distributor of electricity, is the rising tampering. The electric meters were implemented in incidences of illegal tampering of electric meters, a LoRaWAN network to wirelessly transmit watt-hour ‘jumpers,’ and stealing of electric conductors. These acts data and theft notifications, which were displayed on of pilferage are serious threats to the safety of the the energy consumption monitoring website. community, because these acts result in fires and II. RELATED LITERATURE A. Electricty Pilferage and System Loss in the Philippines Manuscript received October 5, 2018; revised April 10, 2019; accepted May 6, 2019. According to a study [ 1 ] , electric energy is Corresponding author: Maria Criselda B. Loyola (email: mcbloyola frequently stolen by the lower and middle class who are @mcl.edu.ph). motivated by the desire to save money. The same study

categorized the issues of electric power theft to safety, the use of GSM technology provided more advantages economy and society. Power theft is a safety issue, for wireless transmission, as compared with previous because it can cause electrocution and fires. methods. Electricity theft is responsible for economic problems A study in press conducted by Loyola et al. [5] utilized in the electric utility due to revenue loss caused by unpaid microcontrollers, Zigbee and GSM technology to detect electricity of the consumers. The thieves have a tendency power theft and notify the consumer. The study was able to consume more energy, resulting to power quality to detect theft in the form of illegal jumpers and meter problems. An increase in power demand to values greater bypassing, and was successful in notifying the end-user than the transformer-rated power can result to different and the distribution utility via SMS. quality deviations like transformer overload, voltage Using GSM, however, limits the way of notifying the unbalance and steady-state voltage drop on system buses consumer and the distribution utility through Short [2]. Message Service (SMS) only. A comparative study by System loss charges represent the cost of electricity Khare et al. [6] stated that GSM modules offers a lost in the distribution system from the receiving point of limited data service. Using this type of communication private distribution utilities and electric cooperatives to system limits the scalability of the power theft detection the consumer’s metering point as defined by Recoverable and notification system, because it is impractical to System Loss Act of 2016. System loss can be classified install several GSM units to accommodate the data into ‘technical losses or ‘non-technical losses.’ Technical traffic in a large-scale application. losses refer to the losses due to power dissipation that In a study proposed by Tariq et al. [7], wireless occur in electrical system components in transmission sensor networks were used to detect and report the and distribution. Non- technical losses cannot be directly presence of illegal tapping. Their study used a taken into account, because they are losses that are due to Resistive Temperature Sensor (RTS) node to obtain the electricity pilferage and administration errors [3]. real-time measurement of the line resistance of circuit As stated in the R.A. 7832, system loss is charged to branches to detect the presence of illegal tapping. An the consumer, despite it being unconsumed power. The increase in the line resistance is a result of an increase law allows private distribution utilities and rural in the load that is due to illegal users [7]. Data gathered cooperatives to collect the costs of these losses through by the sensor nodes were transmitted through motes the system loss. The Chairman of the Senate Energy linked together to form a network. The network formed Committee Filed Senate Bill 1188 or the Recoverable by the motes connected to each electric meter will System Loss Act in attempt to lower the cost of increase the range of transmission. electricity. This bill seeks to lower the current cap of The integration of ZigBee-based wireless sensor system loss charges as mandated by the R.A. 7832 networks in the smart grid was proposed in [8]. Their from 8.5% to 5% for private distribution utilities and system can measure the consumer’s power consumption, 13% to 10% for electrical cooperatives. The bill also store the data in real-time and display the time of use proposes the exclusion of non-technical system losses values. The same study used ZigBee to collect and in the system loss charge that is passed on to the broadcast data and upload it to the consumer’s personal consumers. Meralco currently stands with 6.5% computer. The smart meter system includes a scheduling system loss charge, which is lower than the prescribed mechanism that allows the consumer to set the time of cap. usage of the electricity. Their smart meter system allows B. Previous Works on the Detection and Notification of the consumer to have control over their energy Electric Power Theft consumption in accordance to the smart grid concept. A study [3] proposed the use of a current monitoring Wireless sensor networks have a wide range of system to detect power theft in low-tension transmission applications in creating a manageable, reliable and lines. It proposed that power theft can be detected flexible smart grid [8]. This study applied the concept of when an imbalance of currents is detected between poles. wireless sensor networks in a residential area for the Current between poles is assumed to be balanced under purpose of smart metering and energy consumption no-load conditions. The presence of load causes a monitoring. The communication protocols the study used change in current between poles. This method utilizes a require multiple sensor nodes in the home and a personal current transformer for measuring currents, a signal computer as its server. conditioner circuit for converting the currents to Another similar study that used wireless sensor voltage, and microcontroller for analyzing measured networks was conducted in [9]. They stated that wireless values. Installing the system in-between poles enables sensor networks have the advantage of better accuracy, easier identification of the location of power theft. A lower power consumption, improved area coverage and similar system was proposed by [4] which utilizes an minimal human intervention. Deployment of multiple energy meter, PIC microcontroller, and GSM technology. sensors can retrieve more accurate data compared to a The energy meter used in this study consisted of a single sensor. current transformer, IR sensor and a magnetic reed A study conducted by Wixted et al. [10] compared switch. This was used to detect the presence of energy the performances of data transmission methods of theft. GSM technology was utilized to notify the end- wireless sensor networks. Their study claimed that user of the presence of theft. Reference [4] stated that cabling, Bluetooth, WiFi, and Zigbee were only suitable

for short-range applications in the span of meters up to system addresses the system loss that is due to inaccurate a hundred meters. Long Range (LoRa) technology is meter readings and meter tampering. suitable for applications that require wide network A study in [13] stated that an energy monitoring and coverage. The use of LoRaWAN technology and control system is essential to maximize energy savings multiple gateways increased network coverage and was and to significantly reduce cost. It proposed real time able to reach into problematic areas. energy consumption monitoring system with the remote Low Power Wide Area Networking (LPWAN) control of switches in a residential or corporate technology is the combination of low data rate and long environment. The project proposed in [13] utilized the range communication. LoRaWAN currently displays its Wi-Fi networks for their communication protocol, increasing popularity for being one of the most because it is affordable, readily available, and successful technologies under LPWAN. It was defined transmission to different modules is secure. The data that LoRaWAN employs LoRa as its physical layer. Its collected from the different points of their system are range could reach up to 15 km in sub-ruban areas and compiled in a MySQL database. It was successful in 2 to 5 km in urban areas. LoRaWAN networks were creating a system that allows smart management of commonly deployed as star networks [11]. energy consumption and reducing energy costs. In Limitations of LoRaWAN technology was identified in addition, these types of sustainable management [11] analytic study of LPWAN technologies. The systems are beneficial to the environment. objective of the study was to provide a more realistic In a study by Son [14], a smart meter was used in understanding of the capabilities and appropriate monitoring the power consumption of selected applications of LoRaWAN technology. The researchers appliances inside a house in Korea. It was emphasized concluded that there were several factors to consider for that energy consumption monitoring was the first step the appropriateness of employing LoRaWAN in a certain in reducing the electricity cost of a house. The energy application. These were the number of end-devices, the consumption module developed in [14] estimated the distance of the end-devices to the gateway, the spreading energy consumption of different appliances based on the factor, and the number of channels of the gateway. The data gathered and stored by the smart meter. study also identified that LoRaWAN is not guaranteed to Before, electric meters were used solely for the be successful in real time monitoring and industrial purpose of electric billings and are ignored by the automation applications, where response time is a critical consumer, because only the representative from the consideration. For metering applications, the study distribution utility knows how to interpret the data concluded that it is appropriate on a case-to-case basis. displayed. Even though digital meters and smart meters The study commended LoRaWAN technology for its are now available in the market, some areas have their success in smart lighting, smart parking, and smart waste meters placed on an elevated, centralized location. They collection. In the case of smart city applications, the cannot monitor their consumption proportional to their number of messages sent per day was limited and latency budget. was not a major issue. LoRaWAN’s key features of wide coverage area and connectivity of a great number of devices were utilized in this application. For large-scale consumers, the use of LoRaWAN technology is more applicable to GSM technology. Low power wide area networking enables a wide area network coverage with minimal human interaction for energy consumption monitoring and theft detection and notification. On the other hand, Zigbee technology can also be considered for its data transmission, but its range is limited as compared to LoRaWAN. C. Energy Consumption Monitoring Beside power theft detection and notification, Islam et al. [12] developed a smart metering system that can Fig. 1. Conceptual framework of electric meter with theft detection, theft notification and energy consumption monitoring using LoRaWAN reliably and accurately monitor power consumption of technology. the end-user. This study addressed the issues of inaccurate meter reading due to human error and power III. METHODOLOGY theft. It utilized the GSM technology, Arduino microcontroller and network-based technologies to The entire system of this project consists of the create this system. Their system allows the consumer to electric meter and the energy consumption monitor. monitor his/her power consumption and billing The electric meter features a theft detection and information through a database. It also allows the notification system, aside from measuring and displaying distribution utility to remotely obtain this information the power consumption of the household. through the transmission of data through a GSM modem Fig. 1 shows that the consumer has access to the installed at both ends, consumer and server. This type of electric meter and the energy consumption monitor via

the internet. Electric meter data and theft notifications The energy consumption monitoring system displays were transmitted to the distribution utility via the data collected from the electric meter. The LoRaWAN technology. The distribution utility owns the consumer can access this database given a unique user server of the database where meter data are collated. A ID and a password that corresponds to their electric user account and admin account in the energy meter. Their electronic bill can also be viewed through consumption monitor are provided for the consumer the database. The distribution utility controls a server and distribution utility respectively. account, which receives the consumption data of all the consumers and notifications of power theft incidents. A. Energy Consumption Monitoring The languages used for the user interface and event The electric meter is the main component of the listening were a combination of JavaScript, Cascading system. The electric meter features a theft detection Style Sheets (CSS), and Hypertext Markup Language and notification function. The electric meter measures (HTML). On the server side of the system, PHP was the energy consumption of the consumer and transmits used as the scripting language, and MySQL was used to the data collected to the distribution utility. construct the database. The theft detection module compares the current The distribution utility receives the data transmitted measured before the meter (ICT) and the current from the electric meters of the consumers through a measured by the meter (IPA). If both current readings gateway. The gateway forwards the data transmissions have values that do not exceed the threshold, there will from the electric meter to a backend system, which is be no succeeding action taken, because theft is not hosted by the things network. Data collated in the present. If there is a disparity in the current readings that things network is then retrieved by an Application exceeds the 9% threshold, theft is assumed to be present. Programming Interface (API) for utilization of the Theft was classified as either ‘illegal tapping’ or ‘meter energy consumption monitoring system. bypassing’ depending on the value of the current. The IMST iC880A LoRaWAN concentrator board A voltage divider circuit and a burden resistor were features long range coverage, high robustness, immunity used to interface the current transformer to the analog against interference, and supports multiple channels and input of the Arduino. Fig. 2 shows the schematic spreading factors in parallel. It requires a Linux host to diagram of the voltage divider circuit for interfacing the run its software. The Linux host used was a Raspberry current transformer to the Arduino Mega 2560. Pi 2 Model B. B. Prototyping Assembly, Data Flow, Control Logic, and Network Topology Fig. 3 shows the physical assembly of the components used in the consumer end device. The components of the electric meter were connected as shown on the schematic diagram. The electric meter’s power source was tapped onto the load side of the electric meter, which is also 230 VAC. The 230 VAC was rectified and reduced to a 12 VDC output by an AC–DC converter. To provide an appropriate DC voltage level for the microcontrollers and LCD display, a DC filter was used to reduce the 12 VDC to a 5 VDC supply.

Fig. 2. Schematic diagram of voltage divider circuit for interfacing a current transformer to an Arduino. The notification alert is sent through the LoRaWAN transceiver to be received by the distribution utility. A warning notification is displayed on the electric meter to alert the consumer of the theft. The LoRaWAN transceiver used is a Microchip LoRaWAN RN2483 assembled breakout board, which operates on 868-MHz. Due to constraints in the number of Arduino Mega 2560’s Pulse Width Modulation pins and a conflict with the library of the TFT display, another microcontroller was used to control the LoRa transceiver. An Arduino Nano communicates with the Arduino Mega 2560 as its slave. I2C communication was implemented to transfer the watt-hour data and theft status readings of the Arduino Mega 2560 to the Arduino Nano. An Arduino Nano was chosen for its compact design and minimal number of pins were required for its connections. Fig. 3. Schematic diagram of electric meter.

Fig. 6. Network topology of the system Fig. 4. Schematic diagram of LoRaWAN gateway. C. Prototyping Assembly, Data Flow, Control Logic, and Network Topology Initial to the assembly of the meter, the current- sensing devices were tested for their accuracy. A digital ammeter was used as reference. The current readings of the power analyzer and current transformer were compared to those from the digital ammeter. Fifteen current readings each were obtained for varying loads. The load was supplied by a constant 220 VAC source. The values of the resistive loads used were 1200-Ω, 600-Ω, 300-Ω, and 100-Ω. Current readings for each load set were interpreted by performing the two- tailed t-test for each current-sensing device. The two-

tailed t-test was chosen as the method of data treatment Fig. 5. Block diagram of the system’s data flow at Consumer E. because of the possibility that the output of the current- Fig. 4 shows the assembly of the distribution utility sensing device may be higher or lower than the end device. The gateway components, as previously reference value and it is appropriate for 30 samples discussed, are connected as shown on the schematic or less. Fifteen samples from each current-sensing diagram. The gateway is powered through the Micro device for each load were taken. Each t-test employed 28 USB port of the Raspberry Pi. The gateway requires an degrees of freedom and 99.9% confidence level. The t- input voltage and current of 5 V and 2 A respectively. test was calculated through the data analysis feature of An external 868-MHz antenna is attached to the MS Excel. concentrator board through a pigtail cable. To test the accuracy of the watt-hour measurement of Fig. 5 shows the data flow of the consumer end the electric meter, a commissioned digital meter was used devices. The microcontroller receives and analyzes data as a reference. The total wattage of the load used was 999 from the power analyzer and the current transformer. The W, which consisted of an LED light bulb, electric fan, data obtained is shown on the display and transmitted via and a hair dryer. The load was constantly supplied from a the LoRaWAN transmitter. 230-VAC source. Measurements were taken after 30 This project uses the star network topology as minutes. shown in Fig. 6. Each meter is considered as a node and The LoRaWAN devices were interconnected through is represented by the circles on Fig. 6. Several nodes are the open-source, decentralized network platform, The connected to the gateway represented by the larger circle, Things Network. Network connectivity was established which is the central point of the nodes. The gateway by obtaining gateway traffic in The Things Network. The uploads the data to the internet then a server, which is nodes transmitted packets to the gateway simultaneously. considered the bus represented by the rectangle. The distances of the nodes from the gateway were 69 m, This network topology was chosen because several star 422 m, and 601 m. networks can be implemented consisting of electric To analyze the reliability of the LoRaWAN meters and a gateway over a certain area. This type of transceiver, the round trip time was recorded. The round network topology can be used in the large scale trip time (RTT) of the system is expressed as the length application of this system. Data is centralized at the of time a packet sent by a source plus the time it took server of the distribution utility. for the data to be acknowledge by the receiving end.

Packet capturing was done through Wireshark Network TABLE II. TWO-TAILED T-TEST RESULTS OF CURRENT READINGS OF THE Analyzer. Twenty trials were taken with each lasting for CURRENT TRANSFORMER FOR A 1200-Ω LOAD 100 seconds. Within the capture period of 100 seconds, Parameter Variable 1 Variable 2 no theft was simulated in the first 30 seconds. Theft was Mean 0.185 0.181333333 Variance 8.25399E-34 4.09524E-05 simulated in the next 30 seconds, and no theft in the last Observations 15 15 40 seconds. The capture filter was set to the Pooled Variance 2.04762E-05 LoRaWAN gateway’s IP address. The node placed in Hypothesized Mean the farthest distance from the LoRaWAN gateway was Difference 0 Df 28 used to obtain the average round-trip time of the network. t Stat 2.219103108 The activity of the gateway was analyzed and graphed P(T<=t) one-tail 0.017378173 through the software. t Critical one-tail 3.408155178 The average round-trip time of the 20 trials was P(T<=t) two-tail 0.034756346 computed and its confidence interval. The confidence t Critical two-tail 3.673906401 interval, CI, was calculated using (1), where X is the B. Accuracy of the Designed Electric Meter mean,  is the standard deviation and n is the number of The designed electric meter’s watt-hour trials. The confidence level used to get the confidence measurement was tested for its accuracy by interval was 95%. comparing its readings with a commissioned digital meter. Table III shows the kilowatt-hour readings of CIXn 1.96 / (1) both the electric meter and commissioned digital meter The kilowatt-hour readings of the electric meter and for a constant time of 30 minutes. The watt-hour commissioned digital meter were compared through the readings of the designed meter were divided by 1000 to two-tailed t-test similar to the testing of accuracy of the match the displayed kilowatt-hour measurement of the current-sensing devices. Thirty samples were taken for commissioned electric meter. the test. For the t-test, 58 degrees of freedom and a TABLE III. COMPARISON OF KILOWATT-HOUR READINGS OF DESIGNED confidence level of 99.9% were used. METER IN REFERENCE TO A COMMISSIONED DIGITAL METER Elapsed time Commissioned Designed Trial IV. RESULTS AND DISCUSSION (in minutes) meter (in kWh) meter (in kWh) A. Accuracy of Current-Sensing Devices 1 30 0.5 0.5248 2 30 0.5 0.5250 The current readings of the digital multimeter 3 30 0.5 0.5245 (DMM), Current Transformer (CT) and power analyzer 4 30 0.5 0.5249 (PA) for a 1,200-Ω load at a constant 220 VAC source 5 30 0.5 0.5248 6 30 0.5 0.5245 are shown in Table I. 7 30 0.5 0.5251 8 30 0.5 0.5251 TABLE I. CURRENT READINGS OF THE DIGITAL MULTIMETER, POWER 9 30 0.5 0.5285 ANALYZER AND CURRENT TRANSFORMER FOR A 1200-Ω Load 10 30 0.5 0.5278 Trial DMM reading PA reading CT reading 11 30 0.5 0.5288 number (in ampere) (in ampere) (in ampere) 12 30 0.5 0.5289 13 30 0.5 0.5286 1 0.185 0.18 0.18 14 30 0.5 0.5287 2 0.185 0.18 0.18 15 30 0.5 0.5290 3 0.185 0.17 0.18 16 30 0.5 0.5291 4 0.185 0.19 0.19 17 30 0.5 0.5248 5 0.185 0.18 0.17 18 30 0.5 0.5243 6 0.185 0.17 0.18 19 30 0.5 0.5281 7 0.185 0.18 0.18 20 30 0.5 0.5253 8 0.185 0.19 0.18 21 30 0.5 0.5239 22 30 0.5 0.5273 9 0.185 0.18 0.18 23 30 0.5 0.5298 10 0.185 0.19 0.19 24 30 0.5 0.5264 11 0.185 0.18 0.18 25 30 0.5 0.5276 12 0.185 0.18 0.18 26 30 0.5 0.5281 13 0.185 0.19 0.18 27 30 0.5 0.5286 14 0.185 0.18 0.19 28 30 0.5 0.5288 15 0.185 0.18 0.17 29 30 0.5 0.5269 30 30 0.5 0.5279 Table II shows the computed values for the two-tailed t-test for the current values of the power analyzer. The The results from Table III were analyzed using two calculated t-value for the two-tailed is less than the t tailed independent t-test to identify if there is a critical value at 99.9% confidence level as shown on statistically significant difference between the kilowatt- Table II. It can be concluded that the current readings of hour readings of the commissioned meter and the the power analyzer is 99.9% as accurate as the digital designed electric meter. Table IV shows the calculated multimeter for a 1200-Ω load. values for the two-tailed t-test. The calculated t-value for

the two-tailed is less than the t critical value at 99.9% confidence level. It can be concluded that there is no statistically significant difference between the kilowatt- hour measurement of the designed meter and commissioned meter. The designed meter is 99.9% as accurate as the commissioned digital meter.

TABLE IV. CALCULATED VALUES FOR THE TWO-TAILED T-TEST OF THE KILOWATT-HOUR READINGS OF THE COMMISSIONED METER AND DESIGNED METER Parameter Variable 1 Variable 2 Mean 0.5 0.526863333 Fig. 9. Theft notifications page of the enermon.tech user interface Variance 0 3.55757E-06 Observations 30 30 Pooled variance 1.77879E-06 Hypothesized mean difference 0 df 58 t Stat -78.00878499 P(T<=t) one-tail 9.88399E-61 t Critical one-tail 3.236795339 P(T<=t) two-tail 1.9768E-60 t Critical two-tail 3.466328795

C. Energy Consumption Monitoring System Fig. 10. Assembled LoRaWAN and gateway set-up Fig. 7 shows the homepage of the energy On Fig. 9, theft notifications page of the user consumption monitoring system website. The website is account is shown. The user is notified of recent theft accessible through the link, enermon.tech. The activity that occurs in their corresponding electric homepage features the log-in page to the energy meter. The number of theft incidents and the type of theft consumption monitoring system. is indicated in this page. D. Network Analysis Shown in Fig. 10 is the interior of the assembled LoRaWAN gateway and the actual location set-up of the antenna. For increased coverage, a longer, high-gain 868- MHz antenna was used. The gateway was connected to the internet through a LAN connection. Height and location of the LoRaWAN gateway were critical factors to be considered. The LoRaWAN gateway was placed on the fifth story of a building inside the school premises. The network capture filter was further set to log Fig. 7. Homepage of enermon.tech the conversation between the electric meter and the The consumption monitor that can be accessed by the LoRaWAN gateway. The IP address of the electric consumer through their user account is shown in Fig. 8. meter was ’13.95.217.18,’ served as the source. The IP Their energy consumption can be viewed in graphical address of the gateway was ‘192.168.254.101’ and and tabular form. The graph can be modified depending served as the destination. The round trip time graph was on the user’s preference. Watt-hour consumption can be extracted from the filters. viewed on a daily, weekly, monthly or date range basis. The network was simulated to capture events of The user also has the option to display the watt-hour electric theft. The average delay in milliseconds was consumption as a bar graph or line graph. The tabular obtained by getting the average roundtrip time for the 20 form displays the date and time the measurement was trials and recorded in Table V. The total average round taken, the kilo-watthour consumption for the day, and trip time of the network was 33.092 milliseconds. theft status. TABLE V. ROUND-TRIP TIME OF EACH TRIAL Trial Average RTT (ms) Trial Average RTT (ms) 1 31.59 11 4.10 2 28.56 12 28.39 3 28.35 13 28.93 4 28.19 14 20.25 5 40.23 15 29.85 6 29.57 16 28.23 7 28.60 17 28.37 8 71.59 18 28.67 9 65.87 19 28.73 10 55.49 20 28.23 Fig. 8. Consumption monitor page of the enermon.tech user Interface

Fig. 11. Average round-trip time graph of the LoRaWAN network

TABLE VI. DISTANCES, DEVICE ADDRESSES, AND DEVICE IDS OF Fig. 13. Application data stream via the things network LORAWAN GATEWAY AND END-DEVICES Distance (m) Device Address Device ID Origin eui-b827ebfffe47dacd LoRaWAN - GW V. CONCLUSIONS AND FUTURE WORK 69 260119C6 mcbloyola A practical solution to the high cost of electricity in 422 260119A5 jbueno 601 26011696 rdeleon the Philippines has been long overdue. This paper addressed the costs due to non-technical system losses in With a mean of 0.033092, standard deviation of support of the Recoverable System Loss Act. The 0.015175 and n equal to 20, the confidence interval objective of the project was to help the distribution calculated was 0.006652928. Shown in Fig. 11 is the utility and consumers in minimizing the incidences of graph of the average round trip time for the 20 trials. The electric pilferage. upper and lower limit of the average round trip time was The electric meter was 100% successful in detecting equal to the confidence interval. This shows that the two types of electric pilferage: illegal tapping and meter round trip time of the network is stable, because of its bypassing; and immediately notifying both the consumer low confidence interval with respect to the mean. and distribution utility of the theft occurrence. The Table VI shows the locations of the LoRaWAN electric meter displays the kilowatt-hour consumption, gateway and end-devices with their respective devices system voltage, and type of theft detected. Using two- addresses and device IDs. The network simulation was tailed independent t-test, the kilowatt-hour measurement ran and the gateway traffic and application data stream of the electric meter was ensured to be 99.9% accurate. received on The Things Network were captured. Wireless transmission of meter data and theft The gateway traffic received when the three nodes notifications were successfully accomplished through the were transmitting data is shown in Fig. 12. The gateway utilization of LoRaWAN. The LoRaWAN gateway was identifies the time, frequency, device address, and able to receive packets as far as 601 m line of sight with payload size of the transmission. The device addresses of an average round trip time of 33.09 milliseconds. This the end-devices listed on Table VI are shown to be showed that theft notifications were received by the received by the gateway. distribution utility and consumer relatively fast. The tree The time, device ID, payload, and decoded payload of network was successfully implemented for three end- the received application data are shown in Fig. 13. The devices and a LoRaWAN gateway. device IDs of the end-devices listed on Table VI are The energy consumption monitor allows the consumer shown to be received in The Things Network application to view their daily, weekly, and monthly energy data stream. This showed that the tree network topology consumption, as well as a summary of their electric bill was established between the LoRaWAN gateway and and theft notifications. This system also reduces the three end-devices. Data from the application stream is incidents of inaccurate meter readings due to human integrated into the energy consumption monitor website. error. If adapted nationwide, the system may contribute to the decrease of the cost of electricity and reduction of damages due to electricity pilferage. The threshold set for illegal tapping can be further examined to become proportional to the load consumed by the household. The impact of the 9% threshold varies per household, therefore setting a threshold proportional to the load consumed by the household will be more appropriate.

ACKNOWLEDGMENT This work had been supported by Mapúa Institute of Technology at Laguna, Malayan Colleges Laguna, City Fig. 12. Gateway traffic via the things network of Cabuyao, Laguna, Philippines.

REFERENCES Maria Criselda B. Loyola received her bachelor’s degree in electrical engineering [1] R. Czechowski and A. M. Kosek, “The most frequent energy from University of Perpetual Help System theft techniques and hazards in present power energy Laguna in 2005 and master’s degree in the consumption,” presented at 2016 Joint Workshop on Cyber- same field from Adamson University, Manila Physical Security and Resilience in Smart Grids, 2016. City in 2009. She is currently serving [2] L. G. Arango, E. Deccache, B. D. Bonatto, H. Arango, P. F. Malayan Colleges Laguna in Cabuyao City, Ribeiro, and P. M. Silveira, “Impact of electricity theft on power Philippines as the Chair of the Electrical quality,” in Proc. 17th Int. Conf. on Harmonics and Quality of Engineering program. She started her Power, Brazil, 2016, pp. 557-562. academic career in 2006 as she became part of the faculty of Adamson University and Mapua University before [3] A. A. Chauhan, “Non-technical losses in power system and joining Malayan Colleges Laguna in 2010. monitoring of electricity theft over low-tension poles,” in Proc. Her team’s work on the performance of super capacitor as energy Second Int. Conf. on Advances in Computing and Communication storage and power source was accepted as conference paper in Progress Engineering, India, 2015, pp. 280-284. in Electronics Engineering, Computer Engineering and Information [4] S. Anusha, M. Madhavi, and R. Hemalatha, “Detection of power Technology and was published just recently (2018) in Journal of theft using GSM,” Int. Journal of Advanced Research Trends in Telecommunication, Electronic and Computer Engineering. In Engineering and Technology, vol. 1, no. 3, pp. 15-17, 2014. addition, her work on electricity theft detection, together with another [5] M. C. B. Loyola, J. J. Apurado, and R. B. Casareno, “Electricity team, was accepted for presentation in International Conference on theft detection and notification system using microcontroller, Computer, Communication and Control Technology last March 2018 ZigBee and GSM technologies,” presented at the 2018 14th Int. and is waiting to be published in the same journal. At present, she is Conf. on Computer, Communication and Control Technology, working on another study on energy management system and Krabi, Thailand, March 20-22, 2018. electricity theft. [6] V. Khare, S. Garg, S. Shukla, and P. Sharma, “Comparative study Asst. Prof. Loyola is an active member of the Institute of Integrated of 1G, 2G, 3G and 4G,” Journal of Engineering, Computers & Electrical Engineers of the Philippines. Applied Sciences, vol. 2, no. 4, pp. 55-63, April 2013. Jhamel B. Bueno was born on September [7] M. Tariq and H. V. Poor, “Real time electricity theft detection in 19, 1996. He obtained his Bachelor of microgrids through wireless sensor networks,” in Proc. IEEE Science in Electrical Engineering degree SENSORS, Orlando, FL, 2017. from Malayan Colleges Laguna, Cabuyao [8] M. Burunkaya and T. Pars, “A smart meter design and City, Philippines in April 2018. In his final implementation using ZigBee based wireless sensor network in study year, he represented the institution in a smart grid,” in Proc. 4th Int. Conf. on Electrical and Electronic regional Mathematics quiz bee and his team Engineering, Turkey, 2017, pp. 158-162. ranked second, making them qualified for the [9] J. Patil and A. Kulkarni, “Wireless sensor network using flood national level. He used to be a student leader monitoring,” Int. Journal of Computer Science and Mobile serving the supreme student council of the Computing, vol. 2, no. 3, pp. 297-302, November 2013. Mapua Institute of Technology at Laguna for two academic years. He [10] A. J. Wixted, P. Kinnaird, H. Larijani, A. Tait, A. Ahmadinia, and finished his secondary education at Tabaco National High School, N. Strachan, “Evaluation of LoRa and LoRaWAN for wireless Albay under the Engineering and Science Education Program. sensor networks,” in Proc. IEEE SENSORS, Orlando, FL, 2017. Roshmeir D. De Leon was born on [11] F. Adelantado, X. Vilajosana, P. Tuset-Peiro, B. Martinez, J. November 10, 1995. She attended secondary Melia-Segui, and T. Watteyne, “Understanding the limits of school in South Mansfield College in LoRaWAN,” IEEE Communications Magazine, vol. 55, no. 9, pp. Muntinlupa City, Philippines. She graduated 34-40, September 2017. with the awards of 1st Honorable Mention, [12] M. S. Islam and M. S. R. Bhuiyan, “Design and implementation of Best in Thesis and Thesis Defense. She first remotely located energy meter monitoring with load control and attended De La Salle University-Manila as an mobile billing system through GSM,” in Proc. Int. Conf. on Electronics Engineering student in 2012 Electrical, Computer and Communication Engineering, before transferring to Malayan Colleges Bangladesh, 2017, pp. 158-163. Laguna in 2015 under the Electrical [13] S. Z. Sanchez, R. M. Fernandez-Canti, J. A. Lazara, I. O. Gomez, Engineering program. She became the president of the Institute of and J. A. A. Navarro, “Monitoring and remote control of energy Integrated Electrical Engineers - Malayan Colleges Laguna Student consumption by WiFi networks,” in Proc. IEEE 11th Int. Multi-Conf. Chapter (IIEE - MCL SC) for the school year 2017-2018. She is also a on Systems, Signals & Devices, 2014, pp. 1-5. writer for Malayan Colleges Laguna’s official student publication, [14] S. Y. Son, “Home electricity consumption monitoring Kamalayan. enhancement using smart device status information,” Int. Journal of Smart Home, vol. 9, no. 10, pp. 189-196, 2015.